Hm. Wonder what, if any, effect this will have on lithography node shrinks going forward. I don't know much about this part of the industry, but ASML's near-monopoly can't be good for anyone wanting innovation, no matter how much their customers push them to improve their equipment. If there's only one company capable of delivering the most crucial equipment for producing every single high-end microchip in the world, that's rather scary.
This is exactly why the roll-out of EUV is so mixed across the top 3 players. There was a limit on available equipment because only ASML was making a viable platform for HVM in 2017/2018, with at most 1 unit delivered per month.
Samsung bought the lion's share of the platforms in 2018, and only now are TSMC and Intel able to acquire the number of scanners needed to actually USE EUV in their processes.
Risk was certainly a consideration by all companies when deciding when to insert EUV, but there was just a hard supply limit as well.
source: colleague who went to work for ASML in 2017
Other companies have tried different methods of achieving node shrinks. It's not like EUV was the only attempt, and even just 1 year ago you had people in these forums saying EUV wasn't going to work. 3 years ago you had people outside these forums saying EUV wasn't going to work.
The development of EUV was subsidized by Intel, Samsung, etc. Intel chose not to employ EUV. They probably had the option before or alongside Samsung. Intel, however, expected their roadmap to look differently when they were making their decisions. It was less than 2 years ago, if I remember, when Intel was deciding on EUV use for their 7 nm node and and decided against it because they were worried it wouldn't be ready in time. I am sure they now wish they had made the opposite decision. They are adjusting to insert EUV. EUV wafers per hour and availability have improved faster than expected and Intel's nodes have been pushed back.
From the sound of it by ASML's CEO, what is holding back EUV tool production is the optics systems at Carl Zeiss. Perhaps it won't be much of an issue beyond 2019, though, because there are now only 3 foundries that are currently making the next-generation node shrink (after GlobalFoundries dropped out) and EUV isn't cost effective for memory, yet.
In any case, DUV is still usable, it just gets expensive. And DUV systems will be sold for a long time. ASML shares productivity increases from tools equally with their partners and that won't change because of a switch to EUV. ASML aren't not going to squeeze customers over EUV when it would end up costing them their DUV market. ASML also won't use EUV for pressure to get companies to adopt their DUV tools since ASML are already outcompeting their rivals in DUV anyway. If they start to do that then I am sure regulators will be quick to jump on board. I don't think it's scary.
EUV has been very long coming. ASML was the *only* company that took the risk and spent the money in R&D to develop it. They are now reaping the rewards. CANON and Nikon decided not to spend the ridiculous amount of money necessary to reach production level and only be secondary sources once the technology hit mainstream. ASML almost died from that bet. They needed direct investment from all their possible customers (intel, TSMC, Samsung) to stay afloat during that period. https://www.barrons.com/articles/intel-follows-sam...
Having a monopoly on a product is not against the law. It's abuse of that monopoly which is illegal. So far, ASML has not abused their "power". They need their customers as much as they need ASML.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
6 Comments
Back to Article
Valantar - Saturday, January 26, 2019 - link
Hm. Wonder what, if any, effect this will have on lithography node shrinks going forward. I don't know much about this part of the industry, but ASML's near-monopoly can't be good for anyone wanting innovation, no matter how much their customers push them to improve their equipment. If there's only one company capable of delivering the most crucial equipment for producing every single high-end microchip in the world, that's rather scary.FullmetalTitan - Saturday, January 26, 2019 - link
This is exactly why the roll-out of EUV is so mixed across the top 3 players. There was a limit on available equipment because only ASML was making a viable platform for HVM in 2017/2018, with at most 1 unit delivered per month.Samsung bought the lion's share of the platforms in 2018, and only now are TSMC and Intel able to acquire the number of scanners needed to actually USE EUV in their processes.
Risk was certainly a consideration by all companies when deciding when to insert EUV, but there was just a hard supply limit as well.
source: colleague who went to work for ASML in 2017
Yojimbo - Saturday, January 26, 2019 - link
Other companies have tried different methods of achieving node shrinks. It's not like EUV was the only attempt, and even just 1 year ago you had people in these forums saying EUV wasn't going to work. 3 years ago you had people outside these forums saying EUV wasn't going to work.The development of EUV was subsidized by Intel, Samsung, etc. Intel chose not to employ EUV. They probably had the option before or alongside Samsung. Intel, however, expected their roadmap to look differently when they were making their decisions. It was less than 2 years ago, if I remember, when Intel was deciding on EUV use for their 7 nm node and and decided against it because they were worried it wouldn't be ready in time. I am sure they now wish they had made the opposite decision. They are adjusting to insert EUV. EUV wafers per hour and availability have improved faster than expected and Intel's nodes have been pushed back.
From the sound of it by ASML's CEO, what is holding back EUV tool production is the optics systems at Carl Zeiss. Perhaps it won't be much of an issue beyond 2019, though, because there are now only 3 foundries that are currently making the next-generation node shrink (after GlobalFoundries dropped out) and EUV isn't cost effective for memory, yet.
In any case, DUV is still usable, it just gets expensive. And DUV systems will be sold for a long time. ASML shares productivity increases from tools equally with their partners and that won't change because of a switch to EUV. ASML aren't not going to squeeze customers over EUV when it would end up costing them their DUV market. ASML also won't use EUV for pressure to get companies to adopt their DUV tools since ASML are already outcompeting their rivals in DUV anyway. If they start to do that then I am sure regulators will be quick to jump on board. I don't think it's scary.
frenchy_2001 - Tuesday, January 29, 2019 - link
EUV has been very long coming.ASML was the *only* company that took the risk and spent the money in R&D to develop it. They are now reaping the rewards. CANON and Nikon decided not to spend the ridiculous amount of money necessary to reach production level and only be secondary sources once the technology hit mainstream.
ASML almost died from that bet. They needed direct investment from all their possible customers (intel, TSMC, Samsung) to stay afloat during that period.
https://www.barrons.com/articles/intel-follows-sam...
Having a monopoly on a product is not against the law. It's abuse of that monopoly which is illegal.
So far, ASML has not abused their "power". They need their customers as much as they need ASML.